How In a neuromorphic computing lab in Campinas, Brazil, generates massive data streams that reflect the next frontier in AI efficiency

What if a lab in Campinas, Italy, runs a spiking neural network with 15,000 neurons—each spiking roughly 25 times per second—while storing rich synaptic data? This isn’t science fiction: real neuromorphic systems are pushing boundaries, generating terabytes of data under sustained operation. As interest in brain-inspired computing grows, understanding the scale of data these networks produce reveals compelling insights into both technological advancement and digital infrastructure demands.

Why This Data Generation Matters Today

Understanding the Context

Across the U.S. and globally, neuromorphic computing is emerging as a key solution to energy-intensive AI workloads. Labs like the one in Campinas are pioneering hardware that mimics neural behavior, enabling faster, more efficient processing of complex patterns. With 15,000 neurons firing at 25 spikes per second—equivalent to millions of rapid micro-events per minute—each producing 120 bytes of synaptic activity, the raw data volume becomes a measurable indicator of real-world processing scale.

This surge in data output underscores broader trends: accelerating demand for low-power, high-efficiency computing systems; growing investment in next-generation AI hardware; and the urgent need to manage massive, fast-moving datasets within evolving digital limits.

How Much Data Generated in 10 Minutes? A Clear Breakdown

Each neuron fires an average of 25 times per second. In 10 minutes—600 seconds—one neuron produces:
25 spikes/sec × 600 sec = 15,000 spike events.

Key Insights

With 15,000 neurons in the network, the total spike events are:
15,000 neurons × 15,000 spike events = 225,000,000 spike events.

Each spike records 120 bytes of synaptic activity data, so total bytes per 10 minutes:
225,000,000 spikes × 120 bytes = 27,000,000,000 bytes.

Convert bytes to terabytes:
27,000,000,000 bytes = 27 billion bytes.
Since 1 terabyte = 1,000,000,000,000 bytes, this equals:
27,000,000,000 ÷ 1,000,000,000,000 = 0.027 terabytes.

So, in just 10 minutes, the lab generates approximately 0.027 TB—about 27 gigabytes—of synaptic data. While modest per single session, real systems run continuous cycles, compounding massive data volumes over time.

Common Questions About Data Generation in Neuromorphic Labs

Final Thoughts

Q: How much data do actual neuromorphic systems generate beyond this example?
A: Real-world deployments use larger, more complex networks, often running multi-stage processing. Data rates scale nonlinearly, with systems generating terabytes per day to support learning, simulation, and real-time inference.

Q: Why is 120 bytes per spike event standard?
A: Synaptic data includes timing, strength, and state changes—sufficient for modeling plasticity but compact enough to sustain real-time operation in large-scale networks.

**Q: How do these numbers affect